• Energy Research

Nowadays, increased flood risk is recognized as one of the most significant threats in most parts of the world, with recurring severe flooding events causing significant property and human life losses. This has entailed public debates on both the apparent increased frequency of extreme events and the perceived increases in rainfall intensities within climate changing scenarios. In this work, a stationary vs. Non-Stationary Analysis of annual extreme rainfall was performed with reference to the case studies of the African cities of Dar Es Salaam (TZ) and Addis Ababa (ET). For Dar Es Salaam (TZ) a dataset of 53 years (1958–2010) of maximum daily rainfall records (24 h) was analysed, whereas a 47-year time series (1964–2010) was taken into account for Addis Ababa (ET). Both gauge stations rainfall data were suitably fitted by Extreme Value Distribution (EVD) models. Inference models using the Maximum Likelihood Estimation (MLE) and the Bayesian approach were applied on EVD considering their impact on the shape parameter and the confidence interval width. A comparison between a Non-Stationary regression and a Stationary model was also performed. On this matter, the two time series did not show any Non-Stationary effect. The results achieved under the CLUVA (Climatic Change and Urban Vulnerability in Africa) EU project by the Euro-Mediterranean Centre for Climate Change (CMCC) (with 1 km downscaling) for the IPCC RCP8.5 climatological scenario were also applied to forecast the analysis until 2050 (93 years for Dar Es Salaam TZ and 86 years for Addis Ababa ET). Over the long term, the process seemed to be Non-Stationary for both series. Moreover, with reference to a 100-year return period, the IDF (Intensity-Duration-Frequency) curves of the two case-studies were estimated by applying the Maximum Likelihood Estimation (MLE) approach, as a function of confidence intervals of 2.5% and 97.5% quantiles. The results showed the dependence of Non-Stationary effects of climate change to be conveniently accounted for engineering design and management.

Abstract The increasing impact of flooding urges more effective flood management strategies to guarantee sustainable ecosystem development. Recent catastrophes underline the importance of avoiding local flood management, but characterizing large scale basin wide approaches for systemic flood risk management. Here we introduce an information-theoretic Portfolio Decision Model (iPDM) for the optimization of a systemic ecosystem value at the basin scale by evaluating all potential flood risk mitigation plans. iPDM calculates the ecosystem value predicted by all feasible combinations of flood control structures (FCS) considering environmental, social and economical asset criteria. A multi-criteria decision analytical model evaluates the benefits of all FCS portfolios at the basin scale weighted by stakeholder preferences for assets’ criteria as ecosystem services. The risk model is based on a maximum entropy model (MaxEnt) that predicts the flood susceptibility, the risk of floods based on the exceedance probability distribution, and its most important drivers. Information theoretic global sensitivity and uncertainty analysis is used to select the simplest and most accurate model based on a flood return period. A stochastic optimization algorithm optimizes the ecosystem value constrained to the budget available and provides Pareto frontiers of optimal FCS plans for any budget level. Pareto optimal solutions maximize FCS diversity and minimize the criticality of floods manifested by the scaling exponent of the Pareto distribution of flood size that links management and hydrogeomorphological patterns. The proposed model is tested on the 17,000 km2 Tiber river basin in Italy. iPDM allows stakeholders to identify optimal FCS plans in river basins for a comprehensive evaluation of flood effects under future ecosystem trajectories.